In the recent past I noticed multiple times that contributors use the generic 'noreply' mail that Github provides for their users as the mail associated with their Eclipse account and for submitting contributions (which pass the ECA).
The mail usually has the form: 12345678+myGitHubUsername@users.noreply.github.com
Since it is a 'noreply' mail I assume it's not possible to really contact one through that mail and e.g. if that contributor becomes a committer mailing lists will then probably also not work.
Therefore I wonder if using a noreply-github mail is discouraged and should be forbidden?
I tend to agree. We see these email addresses when folks use the GitHub UI to create commits. At least one concern is whether or not a patch can be adequately tested when it is created in that manner.
Having said all that... I believe that these pass the ECA checker because it understands these email addresses and know how to track back from the GitHub ID to the corresponding Eclipse Foundation Account. If they are passing the ECA checker, we can be sure that they do have an Eclipse Foundation Account with a real email address attached at the EMO can use to contact somebody if necessary. So, we're confident that we can trace these commits back to the author.
If they're a contributor, you can communicate with the contributor via the pull request. If the vibe doesn't feel right, the EMO/IP Team can help committers sort through it.
IMHO, this sort of restriction would be a big change. We need to have very broad support from the committer community before we consider doing this.
I believe that these pass the ECA checker because it understands these email addresses and know how to track back from the GitHub ID to the corresponding Eclipse Foundation Account.
This is correct! There are 2 formats of noreply email (legacy and modern) that both include the Github ID. We can map to the user account as long as it matches the Github handle set in the user account.
It would be unfortunate to prohibit the use of the GitHub UI for commits. You may be correct about testing when the commit is deeply code-related, but a quick documentation fix or something similarly innocuous should absolutely be allowed.
Even when I use non-noreply - if in EF project my PR (after ECA validates my real address used in original commit) gets accepted and merged with "Squash & Merge" or "Rebase & Merge" - it goes to 4% bin.
@wbeaton So long as the ECA validation works, I think this is a non-issue. IMO, the validation is the key component here, not that the commit is associated with an email address someone can cut and paste into an email client and speak with the author. Comments should be kept to the tooling for something like this.
@wbeaton I also think it should be permitted and explicitly supported:
We can always communicate through the official channels (PR, Issue, Discussion, ...) if any direct communication is desired one can ask the person to contact one.
If one wants to register to an mailinglist it needs to use the eclipse account anyways
If the person later wants to become an committer it only works by providing the "official" mail address
I would even go further and make the whole contribution process easier for committers, e.g when I open an issue on Apache and it is only a few lines I can tick a checkbox:
Others offer some automated one-click-workflow see for example here:
so whatever we can do to make the first contribution easier the higher the chance people do it and maybe then contribute more, requiring them to expose their real mail-address "just in case" seem not to fit to such goal.
Since open-source projects and especially Eclipse projects care a lot about transparency I think it's also fair to ask a contributor to be transparent about their real identity. I know everything can be faked, but a completely hidden identity is even less trustworthy to me. Also I think it would be bad for a community if this is used often and we just have anonymous Github accounts that contribute.
And there are also technical aspects that IMO speak against it:
What happens if one day Github goes away or at least if EF-projects migrate away from GH, then it will become very difficult to identify the real person that did a change from the Git history.
As the git history to be the source that I think will last the longest I think we should have as much information in it as possible (this also applies to commit messages).
Furthermore as a committer and PL I'm really not willing to put extra effort into finding out how to contact a person that makes significant contribution, for example to ask if there is interest to become a committer. If one uses a no-reply mail, well then I respect that and doesn't contact that person.
Not every email address is set in stone and for someone who is actively contribution their github id might last longer than email address. Also an e-mail address itself does not mean the user can be contacted there reliable.
@hwellmannwr6 I agree with the general sentiments you have on this topic and I think that any committer / PL that sees a commit come through from a non-obvious identity (regardless of whether it passes ECA) has an obligation to raise questions on it.
However, I want to ensure that we maintain lower barriers for people providing small fixes as just their GitHub identity.
If one uses a no-reply mail, well then I respect that and doesn't contact that person.
+1. If they are using no-reply emails then it is hard to imagine them fully participating in responsibilities of a committer (dev list, voting, dev calls, etc).
And there are also technical aspects that IMO speak against it: What happens if one day Github goes away or at least if EF-projects migrate away from GH, then it will become very difficult to identify the real person that did a change from the Git history.
AFAIU we do not rely on GitHub's presence to identify real identity. The noreply email address is looked up in EF's ECA database for the full details, including real name + real email address if needed. I don't think it is any easier or harder than identifying other ways people disappear (e.g. not updating email address with EF, the old CVS user ids, etc)
From the EF's perspective, we believe that we have what we need with regard to record keeping.
I completely support this position:
If one uses a no-reply mail, well then I respect that and doesn't contact that person.
+1
FWIW, if you feel that a contributor is not representing themselves in good faith, you can reach out to emo-ip-team@eclipse-foundation.org for advice and assistance.
I've set the due date to March 13. I'll ask the Architecture Council to help me decide how to resolve this issue during the meeting that we have scheduled on that date.
@wbeaton - I am not able to attend the AC meeting on that date - I hope that my previous comments on this subject are sufficient explanation of my position that you can resolve this during the AC meeting.
So... we got all bogged down on a rather lively discussion about open source project lifecycle and SBOMs, and didn't get to resolving this on the call. I'll move this to the mailing list.
EMO is satisfied that the documentation requirements in the IP Policy are met when contributors (who must have signed the ECA) provide us with commits using their GitHub "noreply" email.
Based on the responses that we've received from the Eclipse Architecture Council, the EMO will not escalate this matter at this time.